say mostly blue state governors, pennsylvania falls into the category as well, have looked at these shuttered facilities and been like, we closed these down when our constituents, our coalition was very worried about nuclear power. but now, we have to supply new electricity to the grid, supply additional electricity to the grid, and our coalition is more worried about climate change. nuclear power, see what you will, is zero carbon. maybe we should look at bring some of these plants back. california has caused the decommissioning of diablo canyon. michigan and pennsylvania are trying to repower shuttered nuclear power plants, turning them back on. here in new york, we have an unusual situation. a big player in our state level electricity market is a state

agency created around the new >> we take you live to capitol hill for a senate hearing on the chinese challenge to american leadership in artificial intelligence. you're watching c-span3. >> i want to go over a few ground rules. there will be zero tolerance for protests or any efforts to communicate with witnesses or any of the senators here. if you choose to do so, to disrupt this hearing, you will be arrested immediately and barred from the committee for one year. we invite the public to attend, but we also have important business to attend to. with that said, we welcome everyone here today and thank our witnesses for agreeing to testify. we will begin with opening statements from myself, followed by the ranking member, senator coons, and then we'll move to five-minute rounds of questions. with that, i will go ahead and begin. almost 70 years ago, sputnik

launched the greatest technological competition in u.s. history. the space race with the soviet union transformed global science and reshaped the cold war. today we face a similar contest, this time with communist china, and even higher stakes. artificial intelligence will revolutionize daily life and its military uses will shape the global balance of power. beijing is racing to fuse civilian a.i. with its military to seize the next revolution in military affairs. however, unlike the moon landing, the finish line in the a.i. race is far less clear. for the u.s., it may be achieving artificial general intelligence, human level or greater machine cognition. communist china, by contrast, is focused on rapidly infusing a.i. across industry, manufacturing, robotics, and smart cities. these diverging visions were made clear this summer when both

the u.s. and communist china released new a.i. strategies. the trump administration's a.i. action plan centers on openness, markets, and the arms. cutting red shape, building data centers and other a.i. infrastructure, exporting our full a.i. technology stack, including hardware, models, software, applications, and standards to allies and partners, countering beijing in international forum, and enforcing export controls. beijing's a.i. strategy championed state control, sovereignty, and sustained high-level diplomacy to shape the global environment. what's at stake is simple. a u.s.-led future that benefits the free world, or a china system that shapes in line with to be taken values. the risk could not be higher. this race will be won by who attracts the best talent wields the best chips and trains the best algorithms.

america's early lead rests in large part on our dominance of global compute power. at the core of compute are advanced a.i. chips, the best of which are made by american companies. tonight beijing acts as these chips are essential. president trump had the foresight in his first administration to restrict specialized semiconductor manufacturing equipment for being sold to communist china. without cutting-edge tools, chinese fans produced a.i. chips that are inferior to ours. today our top chips are roughly five times more powerful than theirs. and we manufacture them at 10 times the annual scale. further controls by the trump and biden administrations on advanced a.i. chips have forced chinese a.i. companies to smuggle u.s. chips to keep up. if we sustain these limits while u.s. firms innovate and diffuse our technology globally, our compute lead will widen exponentially. that's why senator coons and i plan to introduce the safe chips

act. this legislation would put in statute the trump administration's current red lines on what a.i. chips are allowed to be sold to communist china for 30 months to preserve our edge. xi jinping has made clear he has no intention of remaining reliant on american technology. the c.c.p.'s reason doubled down on the rule of technological self-reliance. u.s. and our allies, taiwan, japan and south korea, control nearly every critical mode of the supply chain, which is the most complex in human history. this is a structural advantage that beijing cannot easily overcome. together we want to enforce export controls and close loopholes so xi jinping's strategy fails. to win the a.i. race, we must also confront our vulnerabilities. china has advantages elsewhere. a.i. data centers require massive energy. communist china generates twice the electricity we do and is

rapidly expanding capacity while our grid ages. china leads in domestic a.i., exporting its tech globally, while u.s. adoption lags. it produces 27 million more stem graduates per decade than the u.s. even as we face an a.i. talent shortage. communist china now leads in openai models, accelerating diffusion of its technology, standards and influence. it also dominates critical minerals, the rare earth elements, which are the lifeblood of a.i. tech and hardware. communist china completely controls the sector with near monopolies in magnets, cobalt and lithium. communist china's restrictions on global exports show how it will leverage, how it can use that leverage to dominate the world. and what leverage that creates for it. president trump's recent critical minerals deal with australia and other allies are good first step, but far more is needed.

some argue these advantages mean that communist china will win the a.i. race. they said the same thing during the space race. they misunderstand americans' resolve and ingenuity. they were wrong then, and they were wrong again this time. but success will require the right policies and sustained investment. i look forward to hearing from our excellent panel of witnesses on how to ensure the united states, not communist china, shapes the a.i. ecosystem of the 21st century. i would like now to recognize my colleague and ranking member for this committee, senator koonce. >> thank you, chairman rick he says, and thank you for working so closely together on this critical hearing. ellen enjoying our service together. i'd like to thank our panel of four seasoned and experienced witnesses to help us discuss today how the united states can win the race to ensure global leadership, even global dominance in a.i. a.i., as you've said, chairman,

is the defining technology of the 21st century, out of which so many other significant advances will come, as with the printing press or the steam engine or the electricity developments of previous centuries. a.i. is the transformation technology that will change our world. i'm poised town lock countless opportunities in nearly every sector of the global economy, from healthcare to education, from manufacturing to energy. it also brings risks. it will be dramatically transformational to military and arms, to intelligence, to cybersecurity, and so much else. our dominance in a.i. isn't just essential for our economic and industrial success, but is critical to our national security. so we need to work together to ensure that as the world builds out its a.i. infrastructure, it's built on an american stack. our chips, our cloud infrastructure, and our models. and as a consequence, on a foundation of american values that help ensure free economies and free societies.

we're not alone in this race, as the chairman just said. china has been working overtime to catch up with us for more than a decade, using state-backed policies to invest massively in accelerating a.i. adoption. they have poured money into research, development, deployment. xi set a goal of being able to be the world's leading a.i. power by 2030. i'm thankful he'll hear from this panel this effort hasn't been wholly successful. they do have a commanding energy in energy production, in talent, but is struggling to produce advanced a.i. chips. secretary lutnick testified this spring that china can currently only produce 200,000 advanced a.i. chips per year, while we know american and allied companies produce more than 10 million. this is our sole competitive critical advantage in a.i., and we cannot squander it. so today i look forward to discussing how we should define the u.s. approach to this watershed technology. here's a few foundational

principles i'd like to suggest that i believe my republican colleagues will agree with. maintaining a.i. primacy has to be a central imperative, central to our economic security and geo strategic goals. the world's most powerful super computers and next generation a.i. must be built by u.s. companies in the united states on a foundation of our democratic principles, which requires us to continue to invest in chips, in energy infrastructure, and in e ensuring we are attracting and retaining the world's top talent. we have to export the full a.i. took our allies and partners, so that american a.i. becomes the foundation for the world in this century. and at the same time, we have to deny our adversaries the ability to build their own stacks, so i'm pleased to join with you chairman in introducing the safe chips act. we also need to recognize that china's inability to access computing power is its chief, if not sole, a.i. vulnerability. and we stand a much better chance of winning every part of

the a.i. competition if we deny them access. so i look forward to hearing from our distinguished panel about how we can translate some of these principles into concrete u.s. government actions and to hear your advice on what more we can and should do together. thank you, mr. chairman. >> thank you, senator koonce. now we turn to our witnesses. our first witness today is dr. chris miller. dr. miller is a professor at the fletcher school at tufts university and is a nonresident senior fellow at a.d.i. dr. miller's latest book "chip war" reveals the geopolitical impact of the computer chip. he also advises businesses and asset managers at green mantle, serve as a member of the geopolitics advisory and advises semiconductors and other technological startups. dr. miller, the floor is yours. please keep your remarks to five minutes. >> thank you very much for the

opportunity to address this panel today on the topic of a.i., which i agree will be the most transformative and economic force we see across our economy and society over the coming decades. right now, just three years into the boom in investment in artificial intelligence that was set off by chatgpt, we're only in the early innings of what i expect to be decades of impact on a.i., both in terms of its economic ramifications, increasing productivity, its act to spur the next generation of technological tools, but also its application to military and intelligence use cases. and indeed, right now we're already seeing the early stages of the application of a.i. to defense and military cases, which is why the work of this committee in assessing the importance of a.i. leadership and devising ways to continue american a.i. leadership is, i believe, so important. the use of artificial

intelligence in defense and intelligence is a continuation of a long-running trend of applying the most advanced computing capabilities to military and intelligence applications. this is something that many of the world's major powers have been doing now for many decades, but the new capabilities that a.i. enables will only intensify their reliance on advanced computing. we see this, for example, in war zones today, in russia and ukraine. it's already the case that both sides are using artificial intelligence to sift through intelligence data and identify what a signal and what is noise. it's already the case that militaries and intelligence agencies are relying on systems to guide more, and increasingly with greater degrees of autonomy. that too requires advanced a.i. systems. and it's already the case that a.i. is being used for cyberoffense and for cyberdefense. and all these reasons suggest to me that it's critically important that america does lead in a.i. as indeed the trump

administration has set out in its strategy. i would argue that there are three keyen puts right now to american a.i. leadership. first, the production of electrical power, where we unfortunately lag behind our competitors and adversaries. second, the e computing power. i believe we have a profound edge over china, an edge that's endured over many years and willen tour for at least the foreseeable future. and third, brainpower, where we have a historical advantage that i worry our lead has been slipping over the last several years. so let me take each of those three inputs to a.i. leadership in turn. first, when it comes to electrical power, it's well known that china is leading the world in installation of new electrical power. from every different source. the united states has a grid that is straining to add even small volumes that we're currently adding. a.i. will require vastly more electric electricity to power the data centers that we need, and it's important we do everything we can to provide the power that a.i. data centers

require. but i think it's unlikely we're going to keep up with china's additions of electrical power, which is why it's so important we retain our leads in the two other spheres, computing power and brainpower. when it comes to computing power, the united states, alongside partners in south korea, in taiwan, in the netherlands, leads the world in the design and the manufacture of the chips that make a.i. possible. the key trend in a.i. over the last not just several years, but really the last several decades has been if you want to build and deploy more systems, you need larger volumes of increasingly advanced a.i. chips. this is why the most advanced tools today require vast data centers full of the most capable a.i. chips, and today almost all of the world's a.i. models are trained and deployed on chips that are designed by american companies and produced either in the united states or in most cases in american allies. this is a lead that china has been trying to catch up to now for a very long time. it's already a decade ago in

2014 that china first named semiconductors as a core technology that industrial policies like the made in china 2025 plan tried to catch up in, and it's still the case, 10 years later, that china is still half a decade behind when it comes to the production of advanced semiconductors. for the future, all the estimates we have, private sector, public sector, acted i can't, there's -- academia, there's no estimates that suggest china is going to catch up soon. i think the computing lead america has, along with its allies, is really quite strong. the final and critically important input to a.i. leadership is brainpower. of course, we need lots of smart engineers to build a.i. systems, and many more to deploy them across all aspects of our economy. china's education system is impressive in the number of engineers it turns out, and we've seen over the last couple of years that chinese engineers at companies like deep seek and alibaba are capable of devising models that are almost as good

as those in the united states. i think it's important that the united states also retain its advantages in talent, both in training our own and in attracting the world's best to ensure that we're able to devise the next generation of a.i. systems here, capitalizing on our leadership in both brainpower and computing power. thank you for the opportunity to speak here. >> great, thank you, dr. miller. our second witness is mr. gray allen. mr. allen is a senior advisor with a a.i. center at csis, where he leads a team of experts on a.i. policy, governance, diplomacy, geopolitics, and national security, and hosts the a.i. policy podcast. he has helped shape u.s. and allied a.i. strategy at major global forums, including the g7 and the a.i. safety summit series. with a particular focus on a.i. export controls and the u.s.-china competition. before joining csis, mr. allen served as director of strategy and policy at the department of defense's joint artificial intelligence center.

his earlier experience includes leading market analysis and competitive strategy for blue origin and serving as a corporate strategy consultant in semiconductors, high performance computing, robotics and space technology. mr. allen, the floor is yours as well. >> distinguished members of the committee, thank you so much for the honor of being with you here today. i want to start first by talking about why it matters to u.s. national security to have an a.i. advantage. i think the best way to do this is by an analogy. that is the analogy of computers. because a.i., like computers, are a general purpose technology. so if you go back to 1945, there is one application of digital programmable computers on planet earth, and it is code breaking, and it shortened world war ii by two years. so it starts out with one killer app. but if you fast forward to today, everything that the military does to some greater or

lesser extent depends upon computers. airplanes have computers in them. ships have computers in them. missiles have computers in them. and even the stuff that does not have computers, like bullets or clothing, was designed on computers, and the logistics that got them to the war fighters was managed by computers. this is a general purpose technology that touches everything that we do and enhances everything that we do. and we're in a similar state now with a.i., where there are a number of applications that are already incredibly proactive, ones that i was involved with during the department of my time at the department of defense was using artificial intelligence capabilities to analyze satellite imagery, to analyze drone imagery, because perhaps you don't know, but it turns out that we collect so much imagery that there's no way that we can have human analysts look at it all every single day. and so we use a.i. as a sort of recommendation engine that says this picture is of an empty ocean. this picture has a war ship in, it i recommend you look at this one first. and that kind of incredible

productivity boost that we got all the way back in the late 2010's from a.i. told us that this was going to be the story of computing all over again. where it starts off with incredibly attract and i have useful applications, and slowly but surely filters into everything that we do. and for those who think that a.i. is fundamentally a commercial technology, i want to emphasize the dual use links of this technology are growing stronger of day. if you interact with a large language model, like a chat got, like chatgpt or anthropic or google's gemini, this seems like a harmless chat interface. but the capabilities underpinning that and the types of companies who are good at creating that are also very good at creating very powerful, military and espionage-relevant capabilities. whether that is using your coding agents to help you create business software or using your coding agents to autonomously execute cyberattacks, the

technology stack is almost identical. and this extends even to autonomous vehicles. when i was in the department of defense, the technology stack underpicking large language models and the technology stack underpinning autonomous vehicles was distinct, aside from hardware similarities. but increasingly with the rise of vision language models, we've discovered that it actually helps cars to think in words about the data that their sensors are collecting. so these commercial worlds of autonomous capabilities, large language models, all of that is going to have relevance to the next generation of autonomous aircraft, which the d.o.d. is investing billions of dollars in. >> and we're leaving this to hear from president trump making an announcement on federally backed investment accounts for children. live coverage on c-span3.